1
00:00:03,200 --> 00:00:10,040
The world we live in is awash with
data that comes pouring in
from everywhere around us.

2
00:00:10,040 --> 00:00:14,520
On its own this data
is just noise and confusion.

3
00:00:14,520 --> 00:00:22,520
To make sense of data, to find the
meaning in it, we need the powerful
branch of science - statistics.

4
00:00:22,520 --> 00:00:26,040
Believe me there's nothing
boring about statistics.

5
00:00:26,040 --> 00:00:29,400
Especially not today
when we can make the data sing.

6
00:00:29,400 --> 00:00:33,400
With statistics we can
really make sense of the world.

7
00:00:33,400 --> 00:00:35,040
And there's more.

8
00:00:35,040 --> 00:00:40,440
With statistics, the data deluge, as
it's being called, is leading us

9
00:00:40,440 --> 00:00:46,240
to an ever greater understanding
of life on Earth
and the universe beyond.

10
00:00:46,240 --> 00:00:50,760
And thanks to the incredible
power of today's computers,

11
00:00:50,760 --> 00:00:57,040
it may fundamentally transform the
process of scientific discovery.

12
00:00:57,040 --> 00:01:02,560
I kid you not, statistics is
now the sexiest subject around.

13
00:01:23,000 --> 00:01:25,600
Did you know that there is
one million boats in Sweden?

14
00:01:25,600 --> 00:01:27,960
That's one boat per nine people!

15
00:01:27,960 --> 00:01:31,080
It's the highest number of
boats per person in Europe!

16
00:01:41,080 --> 00:01:45,760
Being a statistician,
you don't like telling
your profession at dinner parties.

17
00:01:45,760 --> 00:01:48,440
But really,
statisticians shouldn't be shy

18
00:01:48,440 --> 00:01:51,320
because everyone wants to
understand what's going on.

19
00:01:51,320 --> 00:01:56,480
And statistics gives us a
perspective on the world we live in

20
00:01:56,480 --> 00:01:59,320
that we can't get in any other way.

21
00:02:03,520 --> 00:02:09,000
Statistics tells us whether
the things we think
and believe are actually true.

22
00:02:19,960 --> 00:02:25,440
And statistics are far more useful
than we usually like to admit.

23
00:02:25,440 --> 00:02:29,600
In the last recession there
was this famous call-in
to a talk radio station.

24
00:02:29,600 --> 00:02:37,280
The man complained, "In times like
this when unemployment rates are up
to 13%, income has fallen by 5%,

25
00:02:37,280 --> 00:02:41,360
"and suicide rates are climbing, and
I get so angry that the government

26
00:02:41,360 --> 00:02:45,520
"is wasting money on things like
collection of statistics."

27
00:02:48,240 --> 00:02:50,360
I'm not officially a statistician.

28
00:02:50,360 --> 00:02:55,280
Strictly speaking,
my field is global health.

29
00:02:58,120 --> 00:03:03,280
But I got really obsessed with stats
when I realised how much people

30
00:03:03,280 --> 00:03:06,240
in Sweden just don't know
about the rest of the world.

31
00:03:06,240 --> 00:03:10,800
I started in our medical
university, Karolinska Institutet,

32
00:03:10,800 --> 00:03:13,960
an undergraduate course
called Global Health.

33
00:03:13,960 --> 00:03:17,360
These students coming to us actually
have the highest grade you can get

34
00:03:17,360 --> 00:03:18,840
in the Swedish college system,

35
00:03:18,840 --> 00:03:22,040
so I thought, "Maybe they know
everything I'm going to teach them."

36
00:03:22,040 --> 00:03:25,680
So I did a pre-test when they came,
and one of the questions

37
00:03:25,680 --> 00:03:28,160
from which I learned a lot
was this one -

38
00:03:28,160 --> 00:03:32,360
which country has the highest
child mortality of these five pairs?

39
00:03:32,360 --> 00:03:34,920
I won't put you at test here,
but it is Turkey

40
00:03:34,920 --> 00:03:37,000
which is highest there, Poland,

41
00:03:37,000 --> 00:03:40,760
Russia, Pakistan, and South Africa.

42
00:03:40,760 --> 00:03:43,080
And these were the result of
the Swedish students.

43
00:03:43,080 --> 00:03:44,760
A 1.8 right answer
out of five possible.

44
00:03:44,760 --> 00:03:49,920
And that means there was a place for
a professor of International Health
and for my course.

45
00:03:49,920 --> 00:03:56,360
But one late night
when I was compiling the report,
I really realised my discovery.

46
00:03:56,360 --> 00:04:01,160
I had shown that Swedish
top students know statistically

47
00:04:01,160 --> 00:04:04,480
significantly less about
the world than the chimpanzees.

48
00:04:06,000 --> 00:04:09,840
Because the chimpanzees
would score half right.

49
00:04:09,840 --> 00:04:12,320
If I gave them two bananas
with Sri Lanka and Turkey,

50
00:04:12,320 --> 00:04:15,600
they would be right
half of the cases,
but the students are not there.

51
00:04:15,600 --> 00:04:20,200
I did also an unethical study
of the professors of
the Karolinska Institutet,

52
00:04:20,200 --> 00:04:25,520
that hands out the Nobel Prize
for medicine, and they are on par
with the chimpanzees there.

53
00:04:28,160 --> 00:04:32,680
Today there's more information
accessible than ever before.

54
00:04:32,680 --> 00:04:35,760
'And I work with my team at
the Gapminder Foundation

55
00:04:35,760 --> 00:04:41,600
'using new tools that help everyone
make sense of the changing world.

56
00:04:41,600 --> 00:04:45,320
'We draw on the masses of data
that's now freely available

57
00:04:45,320 --> 00:04:49,720
'from international institutions
like the UN and the World Bank.

58
00:04:49,720 --> 00:04:53,640
'And it's become my mission to
share the insights

59
00:04:53,640 --> 00:05:00,200
'from this data with anyone who'll
listen, and to reveal how statistics
is nothing to be frightened of.'

60
00:05:02,440 --> 00:05:05,040
I'm going to provide you a view of

61
00:05:05,040 --> 00:05:09,000
the global health situation
across mankind.

62
00:05:09,000 --> 00:05:14,160
And I'm going to do that in
hopefully an enjoyable way,
so relax.

63
00:05:14,160 --> 00:05:17,120
So we did this software
which displays it like this.

64
00:05:17,120 --> 00:05:19,320
Every bubble here is a country -

65
00:05:19,320 --> 00:05:21,320
this is China, this is India.

66
00:05:21,320 --> 00:05:23,560
The size of the bubble
is the population.

67
00:05:23,560 --> 00:05:27,600
I'm going to stage a race between
this sort of yellowish Ford here

68
00:05:27,600 --> 00:05:32,760
and the red Toyota down there
and the brownish Volvo.

69
00:05:32,760 --> 00:05:36,440
The Toyota has a very bad start
down here, and United States,

70
00:05:36,440 --> 00:05:38,280
Ford is going off-road there,

71
00:05:38,280 --> 00:05:40,480
and the Volvo is doing quite fine,
this is the war.

72
00:05:40,480 --> 00:05:43,680
The Toyota got off track, now Toyota
is on the healthier side of Sweden.

73
00:05:43,680 --> 00:05:46,800
That's about where I sold
the Volvo and bought the Toyota.

74
00:05:46,800 --> 00:05:47,960
AUDIENCE LAUGH

75
00:05:47,960 --> 00:05:50,840
This is the great leap forward,
when China fell down.

76
00:05:50,840 --> 00:05:53,080
It was the central planning
by Mao Zedong.

77
00:05:53,080 --> 00:05:56,680
China recovered and said, "Never
more stupid central planning,"

78
00:05:56,680 --> 00:05:57,800
but they went up here.

79
00:05:57,800 --> 00:06:02,560
No, there is one more inequity,
look there - United States

80
00:06:02,560 --> 00:06:07,480
They broke my frame. Washington DC
is so rich over there,

81
00:06:07,480 --> 00:06:13,040
but they are
not as healthy as Kerala in India.
It's quite interesting, isn't it?

82
00:06:13,040 --> 00:06:14,600
LAUGHTER AND APPLAUSE

83
00:06:20,360 --> 00:06:25,520
Welcome to the USA,
world leaders in big cars

84
00:06:25,520 --> 00:06:28,480
and free data.

85
00:06:28,480 --> 00:06:35,880
There are many here who share
my vision of making public data
accessible and useful for everyone.

86
00:06:35,880 --> 00:06:43,440
The city of San Francisco
is in the lead, opening up
its data on everything.

87
00:06:43,440 --> 00:06:47,480
Even the police department is
releasing all its crime reports.

88
00:06:47,480 --> 00:06:50,840
This official
crime data has been turned

89
00:06:50,840 --> 00:06:55,960
into a wonderful interactive map by
two of the city's computer whizzes.

90
00:06:55,960 --> 00:06:58,920
It's community statistics in action.

91
00:07:09,400 --> 00:07:13,320
Crimespotting is
a map of crime reports from the
San Francisco Police Department

92
00:07:13,320 --> 00:07:16,120
showing dots on maps
for citizens to be able to see

93
00:07:16,120 --> 00:07:19,320
patterns of crime around their
neighbourhoods in San Francisco.

94
00:07:19,320 --> 00:07:25,080
The map is not just about individual
crimes but about broader patterns
that show you where crime is

95
00:07:25,080 --> 00:07:27,760
clustered around the city, which
areas have high crime,

96
00:07:27,760 --> 00:07:30,320
and which areas have
relatively low crime.

97
00:07:36,840 --> 00:07:41,440
We're here at the top of
Jones Street on Nob Hill...

98
00:07:42,960 --> 00:07:45,280
..quite a nice neighbourhood.

99
00:07:45,280 --> 00:07:49,600
What the crime maps show us
is the relationship between

100
00:07:49,600 --> 00:07:51,360
topography and crime.

101
00:07:51,360 --> 00:07:54,520
Basically the higher up the hill,
the less crime there is.

102
00:07:56,200 --> 00:07:58,640
You cross over the border

103
00:07:58,640 --> 00:08:00,240
into the flats...

104
00:08:02,800 --> 00:08:09,240
Essentially as soon as you get
into the lower lying areas of Jones
Street the crime just skyrockets.

105
00:08:20,240 --> 00:08:24,160
We're here in
the uptown Tenderloin district.

106
00:08:26,040 --> 00:08:30,320
It's one of the oldest and densest
neighbourhoods in San Francisco.

107
00:08:30,320 --> 00:08:32,400
This is where you go to buy drugs.

108
00:08:32,400 --> 00:08:33,920
Right around here.

109
00:08:37,200 --> 00:08:41,640
We see lots of aggravated assaults,
lots of auto thefts.

110
00:08:41,640 --> 00:08:48,520
Basically a huge part of the crime
that happens in the city happens
in this five or six block radius.

111
00:08:55,640 --> 00:08:58,920
If you've been hearing police sirens
in your neighbourhood,

112
00:08:58,920 --> 00:09:02,000
you can use the map to find out why.

113
00:09:02,000 --> 00:09:05,680
If you're out at night in
an unfamiliar part of town,

114
00:09:05,680 --> 00:09:09,240
you can check the map
for streets to avoid.

115
00:09:09,240 --> 00:09:12,400
If a neighbour gets burgled,
you can see -

116
00:09:12,400 --> 00:09:16,520
is it a one-off or has there been
a spike in local crime?

117
00:09:16,520 --> 00:09:19,480
If you commute through a
neighbourhood and you're worried

118
00:09:19,480 --> 00:09:23,080
about its safety, the fact that we
have the ability to turn off all

119
00:09:23,080 --> 00:09:25,360
the night-time
and middle-of-the-day crimes

120
00:09:25,360 --> 00:09:28,280
and show you just the things that are
happening during the commute,

121
00:09:28,280 --> 00:09:32,880
it is a statistical operation.
But I think to people that are
interacting with the thing

122
00:09:32,880 --> 00:09:38,000
it feels very much more like they're
just sort of browsing a website
or shopping on Amazon.

123
00:09:38,000 --> 00:09:43,520
They're looking at data
and they don't realise
they're doing statistics.

124
00:09:43,520 --> 00:09:47,840
What's most exciting for me
is that public statistics

125
00:09:47,840 --> 00:09:52,640
is making citizens more powerful and
the authorities more accountable.

126
00:10:02,360 --> 00:10:04,760
We have community meetings that
the police attend

127
00:10:04,760 --> 00:10:08,880
and what citizens are
now doing are bringing printouts

128
00:10:08,880 --> 00:10:12,240
of the maps that show where crimes
are taking place,

129
00:10:12,240 --> 00:10:16,120
and they're demanding services
from the police department

130
00:10:16,120 --> 00:10:20,520
and the police department is now
having to change how they police,

131
00:10:20,520 --> 00:10:22,960
how they provide policing services,

132
00:10:22,960 --> 00:10:27,040
because the data is showing
what is working and what is not.

133
00:10:28,560 --> 00:10:31,960
People in San Francisco
are also using public data

134
00:10:31,960 --> 00:10:35,800
to map social inequalities
and see how to improve society.

135
00:10:35,800 --> 00:10:39,720
And the possibilities are endless.

136
00:10:39,720 --> 00:10:43,160
I think our dream
government data analysis project

137
00:10:43,160 --> 00:10:46,240
would really be focused on
live information,

138
00:10:46,240 --> 00:10:51,240
on stuff that was being reported
and pushed out to the world over
the internet as it was happening.

139
00:10:51,240 --> 00:10:55,040
You know, trash pickups,
traffic accidents, buses,

140
00:10:55,040 --> 00:10:57,680
and I think through the kind of
stats-gathering power

141
00:10:57,680 --> 00:11:02,520
of the internet
it's possible to really begin
to see the workings of the city

142
00:11:02,520 --> 00:11:04,760
displayed as a unified interface.

143
00:11:07,320 --> 00:11:09,960
So that's where we are heading.

144
00:11:09,960 --> 00:11:14,760
Towards a world of free data
with all the statistical
insights that come from it,

145
00:11:14,760 --> 00:11:21,800
accessible to everyone, empowering
us as citizens and letting us
hold our rulers to account.

146
00:11:21,800 --> 00:11:26,920
It's a long way from
where statistics began.

147
00:11:26,920 --> 00:11:32,880
Statistics
are essential to us to monitor
our governments and our societies.

148
00:11:32,880 --> 00:11:36,760
But it was our rulers up
there who started

149
00:11:36,760 --> 00:11:40,840
the collection of statistics in the
first place in order to monitor us!

150
00:11:46,880 --> 00:11:51,440
In fact the word 'statistics'
comes from 'the state'.

151
00:11:51,440 --> 00:11:55,600
Modern statistics
began two centuries ago.

152
00:11:55,600 --> 00:11:59,080
Once it got going,
it spread and never stopped.

153
00:11:59,080 --> 00:12:01,640
And guess who was first!

154
00:12:03,280 --> 00:12:07,560
The Chinese have Confucius,
the Italians have da Vinci,

155
00:12:07,560 --> 00:12:10,240
and the British have Shakespeare.

156
00:12:10,240 --> 00:12:12,440
And we have the Tabellverket -

157
00:12:12,440 --> 00:12:16,400
the first ever systematic
collection of statistics!

158
00:12:16,400 --> 00:12:21,640
Since the year 1749
we have collected data

159
00:12:21,640 --> 00:12:26,920
on every birth, marriage and death,
and we are proud of it!

160
00:12:29,120 --> 00:12:32,000
The Tabellverket recorded
information

161
00:12:32,000 --> 00:12:34,040
from every parish in Sweden.

162
00:12:34,040 --> 00:12:39,080
It was a huge quantity of data and
it was the first time any government

163
00:12:39,080 --> 00:12:41,800
could get an accurate
picture of its people.

164
00:12:49,360 --> 00:12:53,360
Sweden had been the greatest
military power in Northern Europe,

165
00:12:53,360 --> 00:12:58,200
but by 1749 our star
was really fading

166
00:12:58,200 --> 00:13:00,920
and other countries
were growing stronger.

167
00:13:00,920 --> 00:13:03,600
At least we were a large power,

168
00:13:03,600 --> 00:13:09,960
thought to have 20 million people,
enough to rival Britain and France.

169
00:13:13,400 --> 00:13:18,160
But we were in for a nasty surprise.

170
00:13:18,160 --> 00:13:20,680
The first analysis
of the Tabellverket

171
00:13:20,680 --> 00:13:24,080
revealed that Sweden
only had two million inhabitants.

172
00:13:24,080 --> 00:13:30,680
Sweden was not just a power
in decline, it also had
a very small population.

173
00:13:30,680 --> 00:13:36,080
The government was horrified
by this finding -
what if the enemy found out?

174
00:13:37,840 --> 00:13:44,560
But the Tabellverket also showed
that many women died in childbirth
and many children died young.

175
00:13:44,560 --> 00:13:48,640
So government took action
to improve the health of the people.

176
00:13:48,640 --> 00:13:52,440
This was the beginning
of modern Sweden.

177
00:13:53,960 --> 00:13:59,000
It took more than 50 years before
the Austrians, Belgians, Danes,

178
00:13:59,000 --> 00:14:02,320
Dutch, French, Germans, Italians

179
00:14:02,320 --> 00:14:08,600
and, finally, the British,
caught up with Sweden
in collecting and using statistics.

180
00:14:24,640 --> 00:14:29,640
It was called political arithmetic.
It was a lovely phrase
that was used for statistics.

181
00:14:29,640 --> 00:14:33,160
Governments could have much more
control and understanding of

182
00:14:33,160 --> 00:14:36,840
the society - how it was working,
how it was developing

183
00:14:36,840 --> 00:14:40,240
and essentially
so they could control it better.

184
00:14:43,360 --> 00:14:47,960
It wasn't just governments who
woke up to the power of statistics.

185
00:14:47,960 --> 00:14:54,600
Right across Europe, 19th
century society went mad for facts.

186
00:14:54,600 --> 00:14:57,600
And, despite its late start,
Britain,

187
00:14:57,600 --> 00:15:01,400
with its Royal Statistical Society
in London,

188
00:15:01,400 --> 00:15:04,000
was soon a statisticians' nirvana.

189
00:15:05,920 --> 00:15:09,960
I love looking at old copies of
the Royal Statistical Society journal

190
00:15:09,960 --> 00:15:11,760
because it's full of such odd stuff.

191
00:15:11,760 --> 00:15:14,840
There's a wonderful paper
from the 1840s

192
00:15:14,840 --> 00:15:19,200
which shows a map of England and
the rates of bastardy in each county.

193
00:15:19,200 --> 00:15:23,560
So you can identify very quickly the
areas with high rates of bastardy.

194
00:15:23,560 --> 00:15:27,240
Being in East Anglia it always
makes me slightly laugh that Norfolk

195
00:15:27,240 --> 00:15:30,720
seems to top the "bastardy league"
in the 1840s.

196
00:15:30,720 --> 00:15:36,800
One of the founders of
the Royal Statistical Society

197
00:15:36,800 --> 00:15:42,120
was the great
Victorian mathematician
and inventor Charles Babbage.

198
00:15:42,120 --> 00:15:50,120
In 1842 he read the latest
poem by an equally great Victorian,
Alfred Tennyson.

199
00:15:50,120 --> 00:15:53,120
Vision of Sin contained the lines:

200
00:15:53,120 --> 00:15:55,800
"Fill the cup, and fill the can

201
00:15:55,800 --> 00:15:58,160
"Have a rouse before the morn

202
00:15:58,160 --> 00:16:03,720
"Every moment dies a man
Every moment one is born."

203
00:16:03,720 --> 00:16:07,360
So keen a statistician was Babbage
that he could not contain himself.

204
00:16:07,360 --> 00:16:09,360
He dashed off a letter to Tennyson

205
00:16:09,360 --> 00:16:12,200
explaining that because of
population growth,

206
00:16:12,200 --> 00:16:13,640
the line should read,

207
00:16:13,640 --> 00:16:18,640
"Every moment dies a man
and one and a 16th is born."

208
00:16:18,640 --> 00:16:22,480
I may add that
the exact figure is 1.067,

209
00:16:22,480 --> 00:16:27,200
but something must be
conceded to the laws of metre.

210
00:16:31,840 --> 00:16:36,640
In the 19th century, scholars all
over Europe did amazing work

211
00:16:36,640 --> 00:16:39,000
in measuring their societies.

212
00:16:39,000 --> 00:16:42,600
They were hoovering up
data on almost everything.

213
00:16:42,600 --> 00:16:46,040
But numbers alone
don't tell you anything.

214
00:16:46,040 --> 00:16:51,320
You have to analyse them,
and that's what makes statistics.

215
00:16:55,760 --> 00:16:59,200
When the first statisticians
began to get to grips with

216
00:16:59,200 --> 00:17:00,400
analysing their data

217
00:17:00,400 --> 00:17:05,760
they seized upon the average, and
they took the average of everything.

218
00:17:09,720 --> 00:17:13,760
What's so great
about an average is that

219
00:17:13,760 --> 00:17:18,640
you can take a whole mass of data
and reduce it to a single number.

220
00:17:21,880 --> 00:17:26,120
And though each of us is unique,
our collective lives produce

221
00:17:26,120 --> 00:17:29,880
averages that can
characterise whole populations.

222
00:17:41,280 --> 00:17:45,360
I looked in my local newspaper
one week and saw a pensioner

223
00:17:45,360 --> 00:17:49,440
had accidentally put her foot on
the accelerator

224
00:17:49,440 --> 00:17:52,560
and crushed her friend
against a wall.

225
00:17:52,560 --> 00:17:56,360
Devastating, hideous,
horrible thing to happen.

226
00:17:56,360 --> 00:18:01,400
And then there was a second one about
a young man who didn't have

227
00:18:01,400 --> 00:18:07,040
a driving licence, was driving a car
under the influence of drugs
and alcohol

228
00:18:07,040 --> 00:18:10,320
and he bashed into a pedestrian
and killed him.

229
00:18:10,320 --> 00:18:15,560
What's remarkable, absolutely
remarkable, if you look at the number

230
00:18:15,560 --> 00:18:22,880
of people who die each year
in traffic crashes,
it's nearly a constant.

231
00:18:22,880 --> 00:18:24,480
What?

232
00:18:24,480 --> 00:18:31,680
All these individual events,
somehow when you sum them all up
there's the same number every year.

233
00:18:31,680 --> 00:18:35,080
And every year, two and a half
times as many men

234
00:18:35,080 --> 00:18:38,880
die in traffic crashes
as women, and it's a constant.

235
00:18:38,880 --> 00:18:44,320
And every year the rate in Belgium
is double the rate in England.

236
00:18:44,320 --> 00:18:47,160
There are these
remarkable regularities.

237
00:18:47,160 --> 00:18:54,800
So that these individual
particular events sum up
into a social phenomenon.

238
00:18:56,560 --> 00:18:58,120
Let's see what Sweden have done.

239
00:18:58,120 --> 00:19:01,560
We used to boast about fast social
progress, that's where we were....

240
00:19:01,560 --> 00:19:05,240
'In my lectures, to tell stories
about the changing world,

241
00:19:05,240 --> 00:19:08,120
'I use the averages
from entire countries,

242
00:19:08,120 --> 00:19:12,160
'whether the average of income,
child mortality, family size

243
00:19:12,160 --> 00:19:13,360
'or carbon output.'

244
00:19:13,360 --> 00:19:16,200
OK, I give you Singapore.
The year I was born,

245
00:19:16,200 --> 00:19:20,560
Singapore had twice the child
mortality of Sweden, the most
tropical country in the world,

246
00:19:20,560 --> 00:19:22,920
a marshland on
the Equator, and here we go.

247
00:19:22,920 --> 00:19:25,160
It took a little time for them
to get independent,

248
00:19:25,160 --> 00:19:27,160
but then they started to grow
their economy,

249
00:19:27,160 --> 00:19:29,840
and they made the social investment,
they got away malaria,

250
00:19:29,840 --> 00:19:33,360
they got a magnificent health system
that beat both US and Sweden.

251
00:19:33,360 --> 00:19:37,600
We never thought it would happen
that they would win over Sweden!

252
00:19:37,600 --> 00:19:40,520
LAUGHTER AND APPLAUSE

253
00:19:40,520 --> 00:19:46,400
But useful as averages are,
they don't tell you the whole story.

254
00:19:48,800 --> 00:19:53,040
On average, Swedish people have
slightly less than two legs.

255
00:19:53,040 --> 00:19:57,560
This is because few people
only have one leg or no legs,

256
00:19:57,560 --> 00:19:59,760
and no-one has three legs.

257
00:19:59,760 --> 00:20:06,240
So almost everybody in Sweden
has more than
the average number of legs.

258
00:20:06,240 --> 00:20:10,840
The variation in data is just
as important as the average.

259
00:20:16,800 --> 00:20:19,400
But how do you get
a handle on variation?

260
00:20:19,400 --> 00:20:23,000
For this, you transform
numbers into shapes.

261
00:20:23,000 --> 00:20:26,320
Let's look again at the number of
adult women in Sweden

262
00:20:26,320 --> 00:20:27,800
for different heights.

263
00:20:27,800 --> 00:20:31,800
Plotting the data as a shape
shows how much their heights

264
00:20:31,800 --> 00:20:36,400
vary from the average
and how wide that variation is.

265
00:20:36,400 --> 00:20:41,520
The shape a set of data makes
is called its distribution.

266
00:20:41,520 --> 00:20:46,080
This is the income distribution
of China, 1970.

267
00:20:46,080 --> 00:20:51,000
This is the income distribution
of the United States, 1970.

268
00:20:51,000 --> 00:20:54,080
Almost no overlap,
and what has happened?

269
00:20:54,080 --> 00:20:56,880
China is growing,
it's not so equal any longer,

270
00:20:56,880 --> 00:21:01,120
and it's appearing here
overlooking the United States.

271
00:21:01,120 --> 00:21:03,480
Almost like a ghost, isn't it?

272
00:21:03,480 --> 00:21:05,160
It's pretty scary.

273
00:21:05,160 --> 00:21:06,680
Rrrr!

274
00:21:06,680 --> 00:21:08,200
LAUGHTER

275
00:21:17,160 --> 00:21:21,280
The statisticians
who first explored distribution

276
00:21:21,280 --> 00:21:25,760
discovered one shape
that turned up again and again.

277
00:21:25,760 --> 00:21:28,120
The Victorian scholar
Francis Galton

278
00:21:28,120 --> 00:21:32,400
was so fascinated he built
a machine that could reproduce it,

279
00:21:32,400 --> 00:21:36,080
and he found it fitted so many
different sets of measurements

280
00:21:36,080 --> 00:21:38,640
that he named it
the normal distribution.

281
00:21:38,640 --> 00:21:45,600
Whether it was people's arm spans,
lung capacities,

282
00:21:45,600 --> 00:21:47,400
or even their exam results,

283
00:21:47,400 --> 00:21:51,360
the normal distribution shape
recurred time and time again.

284
00:21:51,360 --> 00:21:56,360
Other statisticians soon found
many other regular shapes,

285
00:21:56,360 --> 00:22:01,360
each produced by particular kinds
of natural or social processes.

286
00:22:01,360 --> 00:22:05,400
And every statistician
has their favourite.

287
00:22:05,400 --> 00:22:09,280
The Poisson distribution, the Poisson
shape is my favourite distribution.

288
00:22:09,280 --> 00:22:11,120
I think it's an absolute cracker.

289
00:22:15,760 --> 00:22:18,720
The Poisson shape
describes how likely it is

290
00:22:18,720 --> 00:22:21,680
that out-of-the-ordinary things
will happen.

291
00:22:21,680 --> 00:22:24,520
Imagine a London bus stop where
we know that on average

292
00:22:24,520 --> 00:22:26,280
we'll get three buses in an hour.

293
00:22:26,280 --> 00:22:29,280
We won't always get
three buses, of course.

294
00:22:29,280 --> 00:22:33,480
Amazingly, the Poisson shape will
show us the probability

295
00:22:33,480 --> 00:22:37,200
that in any given hour we will get
four, five, or six buses,

296
00:22:37,200 --> 00:22:39,440
or no buses at all.

297
00:22:40,720 --> 00:22:43,480
The exact shape changes
with the average.

298
00:22:43,480 --> 00:22:46,920
But whether it's how many people
will win the lottery jackpot

299
00:22:46,920 --> 00:22:48,000
each week,

300
00:22:48,000 --> 00:22:51,200
or how many people will phone
a call centre each minute,

301
00:22:51,200 --> 00:22:54,120
the Poisson shape
will give the probabilities.

302
00:22:57,240 --> 00:23:01,240
The wonderful example where this was
applied to in the late 19th century

303
00:23:01,240 --> 00:23:04,400
was to count each year the number of
Prussian officers,

304
00:23:04,400 --> 00:23:07,520
cavalry officers, who were kicked
to death by their horses.

305
00:23:07,520 --> 00:23:10,240
Now, some years there were none,
some years there were one,

306
00:23:10,240 --> 00:23:13,880
some years there were two,
up to seven, I think,
one particularly bad year.

307
00:23:13,880 --> 00:23:16,680
But with this distribution,
however many years there were

308
00:23:16,680 --> 00:23:19,640
with nought, one, two, three,
four Prussian cavalry officers

309
00:23:19,640 --> 00:23:23,880
kicked to death by their horses,
beautifully obeyed
the Poisson distribution.

310
00:23:42,800 --> 00:23:48,520
So statisticians use shapes to
reveal the patterns in the data.

311
00:23:48,520 --> 00:23:51,000
But we also use images of all kinds

312
00:23:51,000 --> 00:23:54,480
to communicate statistics
to a wider public.

313
00:23:54,480 --> 00:23:57,320
Because if the story in the numbers

314
00:23:57,320 --> 00:24:02,920
is told by a beautiful and clever
image, then everyone understands.

315
00:24:02,920 --> 00:24:09,640
Of the pioneers
of statistical graphics,
my favourite is Florence Nightingale.

316
00:24:24,280 --> 00:24:27,120
There are not many people who realise
that she was known

317
00:24:27,120 --> 00:24:30,520
as a passionate statistician
and not just the Lady of the Lamp.

318
00:24:30,520 --> 00:24:34,720
She said that "to understand God's
thoughts, we must study statistics,

319
00:24:34,720 --> 00:24:37,080
"for these are
the measure of His purpose."

320
00:24:37,080 --> 00:24:40,520
Statistics was for her a religious
duty and moral imperative.

321
00:24:42,080 --> 00:24:45,360
When Florence was nine years old
she started collecting data.

322
00:24:45,360 --> 00:24:48,320
Her data was different
fruits and vegetables she found.

323
00:24:48,320 --> 00:24:50,080
Put them into different tables.

324
00:24:50,080 --> 00:24:52,640
Trying to organise them
in some standard form.

325
00:24:52,640 --> 00:24:55,640
And so we have one of Nightingale's
first statistical tables

326
00:24:55,640 --> 00:24:57,440
at the age of nine.

327
00:25:04,360 --> 00:25:11,440
In the mid 1850s Florence
Nightingale went to the Crimea to
care for British casualties of war.

328
00:25:11,440 --> 00:25:14,400
She was horrified by
what she discovered.

329
00:25:14,400 --> 00:25:19,920
For all the soldiers being blown
to bits on the battlefield,
there were many, many more soldiers

330
00:25:19,920 --> 00:25:25,200
dying from diseases they caught
in the army's filthy hospitals.

331
00:25:25,200 --> 00:25:29,120
So Florence Nightingale
began counting the dead.

332
00:25:29,120 --> 00:25:34,920
For two years she recorded
mortality data in meticulous detail.

333
00:25:34,920 --> 00:25:39,120
When the war was over she persuaded
the government to set up

334
00:25:39,120 --> 00:25:41,360
a Royal Commission of Inquiry,

335
00:25:41,360 --> 00:25:44,680
and gathered her data
in a devastating report.

336
00:25:44,680 --> 00:25:48,480
What has cemented her place in
the statistical history books

337
00:25:48,480 --> 00:25:50,120
are the graphics she used.

338
00:25:50,120 --> 00:25:53,960
And one in particular,
the polar area graph.

339
00:25:53,960 --> 00:25:58,680
For each month of the war,
a huge blue wedge represented

340
00:25:58,680 --> 00:26:02,200
the soldiers who had died
from preventable diseases.

341
00:26:02,200 --> 00:26:05,560
The much smaller red wedges were
deaths from wounds,

342
00:26:05,560 --> 00:26:10,600
and the black wedges were deaths
from accidents and other causes.

343
00:26:10,600 --> 00:26:17,040
Nightingale's graphics were so clear
they were impossible to ignore.

344
00:26:17,040 --> 00:26:19,360
The usual thing around
Florence Nightingale's time

345
00:26:19,360 --> 00:26:23,920
was just to produce tables and
tables of figures - absolutely
really tedious stuff that,

346
00:26:23,920 --> 00:26:26,320
unless you're an absolutely dedicated
statistician,

347
00:26:26,320 --> 00:26:29,240
it's really quite difficult to spot
the patterns quite naturally.

348
00:26:29,240 --> 00:26:33,480
But visualisations, they tell a
story, they tell a story immediately.

349
00:26:33,480 --> 00:26:38,480
And the use of colour
and the use of shape can
really tell a powerful story.

350
00:26:38,480 --> 00:26:41,280
And nowadays of course
we can make things move as well.

351
00:26:41,280 --> 00:26:44,320
Florence Nightingale would have
loved to have played with...

352
00:26:44,320 --> 00:26:48,800
She would have
produced wonderful animations,
I'm absolutely certain of it.

353
00:26:50,880 --> 00:26:54,800
Today, 150 years on,
Nightingale's graphics

354
00:26:54,800 --> 00:26:57,800
are rightly regarded as a classic.

355
00:26:57,800 --> 00:27:00,600
They led to a revolution
in nursing, health care

356
00:27:00,600 --> 00:27:05,880
and hygiene in hospitals worldwide,
which saved innumerable lives.

357
00:27:07,400 --> 00:27:11,040
And statistical graphics has
become an art form of its very own,

358
00:27:11,040 --> 00:27:16,280
led by designers who are
passionate about visualising data.

359
00:27:24,640 --> 00:27:27,120
This is the Billion Pound-O-Gram.

360
00:27:27,120 --> 00:27:29,120
This image arose out of frustration

361
00:27:29,120 --> 00:27:32,280
with the reporting of billion pound
amounts in the media.

362
00:27:32,280 --> 00:27:34,400
£500 billion pounds for this war.

363
00:27:34,400 --> 00:27:36,000
£50 billion for this oil spill.

364
00:27:36,000 --> 00:27:39,440
It doesn't make sense -
the numbers are too enormous
to get your mind round.

365
00:27:39,440 --> 00:27:43,520
So I scraped all this data
from various news sources
and created this diagram.

366
00:27:43,520 --> 00:27:48,680
So the
squares here are scaled according
to the billion pound amounts.

367
00:27:48,680 --> 00:27:51,840
When you see numbers visualised
like this

368
00:27:51,840 --> 00:27:54,240
you start to have a different
relationship with them.

369
00:27:54,240 --> 00:27:56,840
You can start to see the patterns,
and the scale of them.

370
00:27:56,840 --> 00:27:59,600
Here in the corner,
this little square - £37 billion.

371
00:27:59,600 --> 00:28:02,800
This was the predicted cost
of the Iraq war in 2003.

372
00:28:02,800 --> 00:28:06,480
As you can see it's grown
exponentially over the last few years

373
00:28:06,480 --> 00:28:10,560
and the total cost now is
around about £2,500 billion.

374
00:28:10,560 --> 00:28:13,000
It's funny because when
you visualise statistics

375
00:28:13,000 --> 00:28:15,360
you understand them,
and when you understand them

376
00:28:15,360 --> 00:28:18,400
you can really start to put things
in perspective.

377
00:28:23,960 --> 00:28:27,880
Visualisation is right at
the heart of my own work too.

378
00:28:27,880 --> 00:28:30,160
I teach global health.

379
00:28:30,160 --> 00:28:33,840
And I know having the data
is not enough -

380
00:28:33,840 --> 00:28:39,160
I have to show it in ways people
both enjoy and understand.

381
00:28:39,160 --> 00:28:42,960
Now I'm going to try something
I've never done before.

382
00:28:42,960 --> 00:28:45,960
Animating the data in real space,

383
00:28:45,960 --> 00:28:50,480
with a bit of technical
assistance from the crew.

384
00:28:50,480 --> 00:28:52,240
So here we go.

385
00:28:52,240 --> 00:28:54,200
First, an axis for health.

386
00:28:54,200 --> 00:28:58,920
Life expectancy
from 25 years to 75 years.

387
00:28:58,920 --> 00:29:01,440
And down here an axis for wealth.

388
00:29:01,440 --> 00:29:06,720
Income per person -
400, 4,000, 40,000.

389
00:29:06,720 --> 00:29:10,480
So down here is poor and sick.

390
00:29:10,480 --> 00:29:14,280
And up here is rich and healthy.

391
00:29:14,280 --> 00:29:18,320
Now I'm going to show you the world

392
00:29:18,320 --> 00:29:21,080
200 years ago, in 1810.

393
00:29:21,080 --> 00:29:22,880
Here come all the countries.

394
00:29:22,880 --> 00:29:26,200
Europe, brown;
Asia, red; Middle East, green;

395
00:29:26,200 --> 00:29:29,440
Africa south of the Sahara,
blue; and the Americas, yellow.

396
00:29:29,440 --> 00:29:33,760
And the size of the country bubble
shows the size of the population.

397
00:29:33,760 --> 00:29:37,560
In 1810, it was pretty crowded
down there, wasn't it?

398
00:29:37,560 --> 00:29:39,760
All countries were sick and poor.

399
00:29:39,760 --> 00:29:43,360
Life expectancy
was below 40 in all countries.

400
00:29:43,360 --> 00:29:48,680
And only UK and the Netherlands were
slightly better off. But not much.

401
00:29:48,680 --> 00:29:52,520
And now I start the world.

402
00:29:52,520 --> 00:29:56,840
The industrial revolution makes
countries in Europe and elsewhere

403
00:29:56,840 --> 00:29:59,040
move away from the rest.

404
00:29:59,040 --> 00:30:02,280
But the colonized countries
in Asia and Africa,

405
00:30:02,280 --> 00:30:04,040
they are stuck down there.

406
00:30:04,040 --> 00:30:08,200
And eventually the Western countries
get healthier and healthier.

407
00:30:08,200 --> 00:30:13,320
And now we slow down to show
the impact of the First World War

408
00:30:13,320 --> 00:30:15,880
and the Spanish flu epidemic.

409
00:30:15,880 --> 00:30:18,320
What a catastrophe!

410
00:30:18,320 --> 00:30:22,640
And now I speed up through
the 1920s and the 1930s and,

411
00:30:22,640 --> 00:30:24,400
in spite of the Great Depression,

412
00:30:24,400 --> 00:30:27,800
Western countries forge on towards
greater wealth and health.

413
00:30:27,800 --> 00:30:29,880
Japan and some others try to follow.

414
00:30:29,880 --> 00:30:32,560
But most countries stay down here.

415
00:30:32,560 --> 00:30:35,640
And after the tragedies
of the Second World War,

416
00:30:35,640 --> 00:30:39,400
we stop a bit to look
at the world in 1948.

417
00:30:39,400 --> 00:30:42,080
1948 was a great year.

418
00:30:42,080 --> 00:30:43,280
The war was over,

419
00:30:43,280 --> 00:30:48,000
Sweden topped the medal table at
the Winter Olympics and I was born.

420
00:30:48,000 --> 00:30:51,280
But the differences between
the countries of the world

421
00:30:51,280 --> 00:30:52,680
was wider than ever.

422
00:30:52,680 --> 00:30:54,960
United States was in the front.

423
00:30:54,960 --> 00:30:56,840
Japan was catching up.

424
00:30:56,840 --> 00:30:58,400
Brazil was way behind,

425
00:30:58,400 --> 00:31:03,040
Iran was getting a little richer
from oil but still had short lives.

426
00:31:03,040 --> 00:31:05,160
And the Asian giants...

427
00:31:05,160 --> 00:31:08,720
China, India, Pakistan, Bangladesh,
and Indonesia,

428
00:31:08,720 --> 00:31:11,360
they were still
poor and sick down here.

429
00:31:11,360 --> 00:31:14,360
But look what was about to happen!
Here we go again.

430
00:31:14,360 --> 00:31:18,640
In my lifetime, former colonies
gained independence and then finally

431
00:31:18,640 --> 00:31:22,640
they started to get healthier
and healthier and healthier.

432
00:31:22,640 --> 00:31:26,080
And in the 1970s, then countries
in Asia and Latin America

433
00:31:26,080 --> 00:31:28,960
started to catch up
with the Western countries.

434
00:31:28,960 --> 00:31:31,240
They became the emerging economies.

435
00:31:31,240 --> 00:31:32,640
Some in Africa follows,

436
00:31:32,640 --> 00:31:36,440
some Africans were stuck in civil
war, and others were hit by HIV.

437
00:31:36,440 --> 00:31:41,840
And now we can see the world
in the most up-to-date statistics.

438
00:31:42,840 --> 00:31:45,480
Most people today
live in the middle.

439
00:31:45,480 --> 00:31:48,080
But there is huge difference
at the same time

440
00:31:48,080 --> 00:31:51,520
between the best-off countries
and the worst-off countries.

441
00:31:51,520 --> 00:31:54,520
And there are also huge
inequalities within countries.

442
00:31:54,520 --> 00:31:59,000
These bubbles show country averages
but I can split them.

443
00:31:59,000 --> 00:32:02,120
Take China. I can split it
into provinces.

444
00:32:02,120 --> 00:32:05,120
There goes Shanghai...

445
00:32:05,120 --> 00:32:08,000
It has the same health
and wealth as Italy today.

446
00:32:08,000 --> 00:32:11,240
And there
is the poor inland province Guizhou,

447
00:32:11,240 --> 00:32:12,680
it is like Pakistan.

448
00:32:12,680 --> 00:32:18,800
And if I split it further, the rural
parts are like Ghana in Africa.

449
00:32:19,800 --> 00:32:23,160
And yet, despite the enormous
disparities today,

450
00:32:23,160 --> 00:32:27,240
we have seen 200 years
of remarkable progress!

451
00:32:27,240 --> 00:32:31,720
That huge historical gap between
the west and the rest is now closing.

452
00:32:31,720 --> 00:32:35,640
We have become an entirely
new, converging world.

453
00:32:35,640 --> 00:32:37,960
And I see a clear trend
into the future.

454
00:32:37,960 --> 00:32:40,840
With aid, trade, green
technology and peace,

455
00:32:40,840 --> 00:32:43,720
it's fully possible
that everyone can make it

456
00:32:43,720 --> 00:32:45,640
to the healthy, wealthy corner.

457
00:32:48,000 --> 00:32:51,360
Well, what you've just seen
in the last few minutes

458
00:32:51,360 --> 00:32:56,520
is a story of 200 countries
shown over 200 years and beyond.

459
00:32:56,520 --> 00:33:00,960
It involved plotting
120,000 numbers.

460
00:33:00,960 --> 00:33:02,560
Pretty neat, huh?

461
00:33:07,960 --> 00:33:13,120
So, with statistics, we can begin
to see things as they really are.

462
00:33:13,120 --> 00:33:18,200
From tables of data to averages,
distributions and visualisations,

463
00:33:18,200 --> 00:33:22,640
statistics gives us a
clear description of the world.

464
00:33:22,640 --> 00:33:28,200
But, with statistics, we can
not only discover WHAT is happening

465
00:33:28,200 --> 00:33:30,520
but also explore WHY,

466
00:33:30,520 --> 00:33:34,480
by using the powerful analytical
method - correlation.

467
00:33:35,480 --> 00:33:38,400
Just looking at one thing at a
time doesn't tell you very much.

468
00:33:38,400 --> 00:33:41,280
You've got to look at the
relationships between things,

469
00:33:41,280 --> 00:33:43,360
how they change,
how they vary together.

470
00:33:43,360 --> 00:33:45,360
That's what correlation is about.

471
00:33:45,360 --> 00:33:48,320
That's how you start trying
to understand the processes

472
00:33:48,320 --> 00:33:50,960
that are really going on
in the world and society.

473
00:33:52,480 --> 00:33:57,000
Most of us today would recognise
that crime correlates to poverty,

474
00:33:57,000 --> 00:34:00,200
that infection correlates
to poor sanitation,

475
00:34:00,200 --> 00:34:02,600
and that knowledge of statistics
correlates

476
00:34:02,600 --> 00:34:05,040
to being great at dancing!

477
00:34:06,560 --> 00:34:10,200
Correlations can be very tricky.

478
00:34:10,200 --> 00:34:12,960
I got a joke about
silly correlations.

479
00:34:12,960 --> 00:34:15,840
There was this American who
was afraid of heart attack.

480
00:34:15,840 --> 00:34:19,920
He found out that
the Japanese ate very little fat

481
00:34:19,920 --> 00:34:22,320
and almost didn't drink wine,

482
00:34:22,320 --> 00:34:25,520
but they had much less
heart attacks than the Americans.

483
00:34:25,520 --> 00:34:28,640
But, on the other hand,
he also found out that the French

484
00:34:28,640 --> 00:34:35,080
eat as much fat as the Americans
and they drink much more wine but
they also have less heart attacks.

485
00:34:35,080 --> 00:34:40,840
So he concluded that what kills you
is speaking English.

486
00:34:40,840 --> 00:34:43,920
# Smoke, smoke,
smoke that cigarette

487
00:34:43,920 --> 00:34:48,000
# Puff, puff, puff and if you
smoke yourself to death... #

488
00:34:48,000 --> 00:34:51,920
The time, the pace,
the cigarette. Weights Tilt.

489
00:34:51,920 --> 00:34:56,200
The best example of a really
ground-breaking correlation

490
00:34:56,200 --> 00:35:01,640
is the link that was established
in the 1950s between
smoking and lung cancer.

491
00:35:01,640 --> 00:35:07,040
Not long after the Second World War,
a British doctor, Richard Doll,

492
00:35:07,040 --> 00:35:11,040
investigated lung cancer patients
in 20 London hospitals.

493
00:35:11,040 --> 00:35:15,400
And he became certain
that the only thing they had
in common was smoking.

494
00:35:15,400 --> 00:35:18,280
So certain,
that he stopped smoking himself.

495
00:35:18,280 --> 00:35:22,160
But other people weren't so sure.

496
00:35:22,160 --> 00:35:25,400
A lot of the discussion
of the early data,

497
00:35:25,400 --> 00:35:29,120
linking smoking to lung cancer, said,
"It's not the smoking, surely,

498
00:35:29,120 --> 00:35:32,600
"that thing we've done all our lives,
that can't be bad for you.

499
00:35:32,600 --> 00:35:35,000
"Maybe it's genes.

500
00:35:35,000 --> 00:35:39,080
"Maybe people who are genetically
predisposed to get lung cancer

501
00:35:39,080 --> 00:35:43,840
"are also genetically
predisposed to smoke."

502
00:35:43,840 --> 00:35:47,360
"Maybe it's not the smoking,
maybe it's air pollution -

503
00:35:47,360 --> 00:35:52,520
"that smokers are somehow
more exposed to air pollution
than non-smokers.

504
00:35:52,520 --> 00:35:56,280
"Maybe it's not smoking,
maybe it's poverty."

505
00:35:56,280 --> 00:36:00,720
So now we've got three alternative
explanations, apart from chance.

506
00:36:02,240 --> 00:36:06,760
To verify his correlation
did imply cause and effect.

507
00:36:06,760 --> 00:36:10,680
Richard Doll created the biggest
statistical study of smoking yet.

508
00:36:10,680 --> 00:36:14,680
He began tracking the lives
of 40,000 British doctors,

509
00:36:14,680 --> 00:36:17,000
some of whom smoked
and some of whom didn't,

510
00:36:17,000 --> 00:36:19,440
and gathered enough data

511
00:36:19,440 --> 00:36:22,000
to correlate the amount
the doctors smoked

512
00:36:22,000 --> 00:36:24,920
with their likelihood
of getting cancer.

513
00:36:24,920 --> 00:36:30,120
Eventually, he not only
showed a correlation between
smoking and lung cancer,

514
00:36:30,120 --> 00:36:35,800
but also a correlation
between stopping smoking
and reducing the risk.

515
00:36:35,800 --> 00:36:37,760
This was science at its best.

516
00:36:39,760 --> 00:36:44,000
What correlations do not replace
is human thought.

517
00:36:44,000 --> 00:36:46,760
You've got to think
about what it means.

518
00:36:46,760 --> 00:36:50,480
What a good scientist does,
if he comes with a correlation,

519
00:36:50,480 --> 00:36:55,960
is try as hard as she or he
possibly can to disprove it,

520
00:36:55,960 --> 00:37:00,200
to break it down, to get rid of it,
to try and refute it.

521
00:37:00,200 --> 00:37:05,440
And if it withstands
all those efforts at demolishing it

522
00:37:05,440 --> 00:37:10,760
and it is still standing up then,
cautiously, you say, "We really
might have something here."

523
00:37:26,720 --> 00:37:32,840
However brilliant the scientist,
data is still the oxygen of science.

524
00:37:32,840 --> 00:37:39,320
The good news is that the more we
have, the more correlations we'll
find, the more theories we'll test,

525
00:37:39,320 --> 00:37:42,240
and the more discoveries
we're likely to make.

526
00:37:46,160 --> 00:37:53,440
And history shows how our total sum
of information grows in huge leaps
as we develop new technologies.

527
00:37:53,440 --> 00:38:00,000
The invention of the
printing press kicked off the first
data and information explosion.

528
00:38:00,000 --> 00:38:06,000
If you piled up all the books that
had been printed by the year 1700,

529
00:38:06,000 --> 00:38:11,200
they would make 60 stacks
each as high as Mount Everest.

530
00:38:12,880 --> 00:38:15,360
Then, starting in the 19th century,

531
00:38:15,360 --> 00:38:19,880
there came a second information
revolution with the telegraph,

532
00:38:19,880 --> 00:38:23,960
gramophone and camera.
And later radio and TV.

533
00:38:23,960 --> 00:38:28,200
The total amount
of information exploded.

534
00:38:28,200 --> 00:38:35,200
And by the 1950s
the information available to us all
had multiplied 6,000 times.

535
00:38:35,200 --> 00:38:41,440
Then, thanks to the computer and
later the internet, we went digital.

536
00:38:41,440 --> 00:38:47,200
And the amount of data we have now
is unimaginably vast.

537
00:38:49,920 --> 00:38:55,080
A single letter printed in a book
is equivalent to a byte of data.

538
00:38:55,080 --> 00:38:58,720
A printed page
equals a kilobyte or two.

539
00:39:01,960 --> 00:39:06,240
Five megabytes is enough for
the complete works of Shakespeare.

540
00:39:08,000 --> 00:39:11,680
10 gigabytes - that's a DVD movie.

541
00:39:16,840 --> 00:39:23,360
Two terabytes
is the tens of millions of photos
added to Facebook every day.

542
00:39:24,880 --> 00:39:32,200
Ten petabytes is the data recorded
every second by the world's
largest particle accelerator.

543
00:39:32,200 --> 00:39:35,800
So much
only a tiny fraction is kept.

544
00:39:35,800 --> 00:39:43,440
Six exabytes is what you'd have
if you sequenced the genomes
of every single person on Earth.

545
00:39:48,680 --> 00:39:50,520
But really, that's nothing.

546
00:39:50,520 --> 00:39:55,080
In 2009, the internet
added up to 500 exabytes.

547
00:39:55,080 --> 00:40:02,120
In 2010, in just one year, that will
double to more than one zettabyte!

548
00:40:06,360 --> 00:40:14,000
Back in the real world, if we
turned all this data into print
it would make 90 stacks of books,

549
00:40:14,000 --> 00:40:18,560
each reaching from here
all the way to the sun!

550
00:40:18,560 --> 00:40:23,600
The data deluge is staggering,
but, with today's computers

551
00:40:23,600 --> 00:40:28,200
and statistics,
I'm confident we can handle it.

552
00:40:28,200 --> 00:40:31,400
When it comes to all the data
on the internet,

553
00:40:31,400 --> 00:40:33,760
the powerhouse
of statistical analysis

554
00:40:33,760 --> 00:40:37,560
is the Silicon Valley giant Google.

555
00:40:44,000 --> 00:40:50,600
The average person over their
lifetime is exposed to about 100
million words of conversation.

556
00:40:50,600 --> 00:40:54,840
And so if you multiple that by the
six billion people on the planet,

557
00:40:54,840 --> 00:40:58,040
that amount of words is about
equal to the number of words

558
00:40:58,040 --> 00:41:01,080
that Google has available
at any one instant in time.

559
00:41:03,480 --> 00:41:08,680
Google's computers hoover up
and file away every document,
web page, and image they can find.

560
00:41:08,680 --> 00:41:14,640
They then hunt for patterns and
correlations in all this data,

561
00:41:14,640 --> 00:41:17,760
doing statistics on a massive scale.

562
00:41:17,760 --> 00:41:25,560
And, for me, Google has one project
that's particularly exciting -
statistical language translation.

563
00:41:25,560 --> 00:41:30,880
We wanted to provide access
to all the web's information,
no matter what language you spoke.

564
00:41:30,880 --> 00:41:33,520
There's just so much information
on the internet,

565
00:41:33,520 --> 00:41:37,880
you couldn't hope to translate it all
by hand into every possible language.

566
00:41:37,880 --> 00:41:41,560
We figured we'd have to be able
to do machine translation.

567
00:41:44,280 --> 00:41:47,360
In the past, programmers
tried to teach their computers

568
00:41:47,360 --> 00:41:53,320
to see each language as a set of
grammatical rules - much like the
way languages are taught at school.

569
00:41:53,320 --> 00:41:58,760
But this didn't work because no set
of rules could capture a language

570
00:41:58,760 --> 00:42:01,480
in all its subtlety and ambiguity.

571
00:42:01,480 --> 00:42:05,840
"Having eaten our lunch
the coach departed."

572
00:42:05,840 --> 00:42:07,920
Well, that's obviously incorrect.

573
00:42:07,920 --> 00:42:12,000
Written like that it would imply
that the coach has eaten the lunch.

574
00:42:12,000 --> 00:42:15,160
It would be far better to say...

575
00:42:15,160 --> 00:42:19,920
"having eaten our lunch
we departed in the coach."

576
00:42:19,920 --> 00:42:26,320
Those rules are helpful and they are
useful most of time, but they don't
turn out to be true all the time.

577
00:42:26,320 --> 00:42:30,320
And the insight of using statistical
machine translation is saying,

578
00:42:30,320 --> 00:42:35,280
"If you've got to have all these
exceptions anyways, maybe you can get
by without having any of the rules.

579
00:42:35,280 --> 00:42:39,480
"Maybe you can treat everything
as an exception." And that's
essentially what we've done.

580
00:42:48,840 --> 00:42:52,640
What the computer is doing when
he's learning how to translate

581
00:42:52,640 --> 00:42:55,160
is to learn correlations
between words

582
00:42:55,160 --> 00:42:57,240
and correlations between phrases.

583
00:42:57,240 --> 00:43:00,840
So we feed the system very large
amounts of data

584
00:43:00,840 --> 00:43:04,720
and then the system is seeing that
a certain word or a certain phrase

585
00:43:04,720 --> 00:43:07,600
correlates very often
to the other language.

586
00:43:09,800 --> 00:43:15,800
Google's website currently
offers translation between
any of 57 different languages.

587
00:43:15,800 --> 00:43:22,680
It does this purely statistically,
having correlated a huge collection
of multilingual texts.

588
00:43:22,680 --> 00:43:25,600
The people that built the system
don't need to know Chinese

589
00:43:25,600 --> 00:43:29,800
in order to build the
Chinese-to-English system,
or they don't need to know Arabic.

590
00:43:29,800 --> 00:43:33,040
But the expertise that's needed is
basically knowledge of statistics,

591
00:43:33,040 --> 00:43:35,840
knowledge of computer science,
knowledge of infrastructure

592
00:43:35,840 --> 00:43:40,880
to build those very large
computational systems
that we are building for doing that.

593
00:43:42,880 --> 00:43:48,360
I hooked up with Google
from my office in Stockholm
to try the translator for myself.

594
00:43:48,360 --> 00:43:51,760
'I will type...
some Swedish sentences.'

595
00:43:51,760 --> 00:43:53,080
OK.

596
00:43:53,080 --> 00:43:55,240
Sveriges...

597
00:43:55,240 --> 00:43:59,280
..guldring i orat.

598
00:44:00,920 --> 00:44:07,400
OK. So it says, "Sweden's finance
minister has a ponytail
and a gold ring in your ear."

599
00:44:07,400 --> 00:44:11,520
I guess it probably means
in his ear. 'That's exactly
correct, it's amazing!

600
00:44:11,520 --> 00:44:15,400
'He comes from the Conservative
party, that's the kind
of Sweden we have today.

601
00:44:15,400 --> 00:44:18,520
'I will type one more sentence.'

602
00:44:18,520 --> 00:44:22,080
'I sitt samkonade...'

603
00:44:22,080 --> 00:44:25,600
partnerskap...

604
00:44:25,600 --> 00:44:28,280
nya biskop.

605
00:44:28,280 --> 00:44:35,200
"In his same-sex partnership
has Stockholm's new bishop
and his partners a three-year son."

606
00:44:35,200 --> 00:44:38,120
It's almost perfect,
there's one important thing -

607
00:44:38,120 --> 00:44:41,800
it's HER,
it's a lesbian partnership.

608
00:44:41,800 --> 00:44:46,760
OK, so those kinds of words his
and her are one of the challenges

609
00:44:46,760 --> 00:44:49,080
in translation
to get really those right.

610
00:44:49,080 --> 00:44:51,920
Especially when it comes
to bishops one can excuse it!

611
00:44:51,920 --> 00:44:53,640
'Right, right.'

612
00:44:53,640 --> 00:44:58,520
I guess more often than not
it would probably be a "his".
'I will write one more sentence.'

613
00:44:58,520 --> 00:45:01,720
Nar Sverige deltar
I olympiader ar malet

614
00:45:01,720 --> 00:45:03,720
'inte att vinna
utan att sla Norge.'

615
00:45:06,400 --> 00:45:11,960
OK. "When Sweden is taking part
in Olympic goal is not
to win but to beat Norway."

616
00:45:11,960 --> 00:45:13,640
'Yes! This is what it is!

617
00:45:13,640 --> 00:45:17,920
'But they are very good
in Winter Olympics, so we
can't make it, but we are trying.'

618
00:45:17,920 --> 00:45:19,960
Ah, very good, very good.

619
00:45:19,960 --> 00:45:24,960
'This is absolutely amazing, you
know, and I was especially impressed

620
00:45:24,960 --> 00:45:30,520
'that it picks up words like
"same-sex partnership"
which are very new to the language."

621
00:45:30,520 --> 00:45:36,920
'The translator is good, but
if they succeed with what's next,
that'll be remarkable.'

622
00:45:36,920 --> 00:45:38,440
One of the exciting possibilities

623
00:45:38,440 --> 00:45:42,720
is combining the machine
translation technology with
the speech recognition technology.

624
00:45:42,720 --> 00:45:45,480
Now, both of these
are statistical in nature.

625
00:45:45,480 --> 00:45:51,360
The machine translation relies
on the statistics of mapping
from one language to another,

626
00:45:51,360 --> 00:45:57,840
and similarly speech recognition
relies on the statistics of mapping
from a sound form to the words.

627
00:45:57,840 --> 00:45:59,520
When we put them together,

628
00:45:59,520 --> 00:46:03,200
now we have the capability
of having instant conversation

629
00:46:03,200 --> 00:46:06,760
between two people
that don't speak a common language.

630
00:46:06,760 --> 00:46:08,680
I can talk to you in my language,

631
00:46:08,680 --> 00:46:11,880
you hear me in your language
and you can answer back.

632
00:46:11,880 --> 00:46:15,000
And in real time we can
make that translation,

633
00:46:15,000 --> 00:46:18,800
we can bring two people together
and allow them to speak.

634
00:46:31,400 --> 00:46:39,040
The internet is just one
of many technologies created
to gather massive amounts of data.

635
00:46:39,040 --> 00:46:43,640
Scientists studying
our earth and our environment

636
00:46:43,640 --> 00:46:47,440
now use an incredible range
of instruments

637
00:46:47,440 --> 00:46:50,920
to measure the processes
of our planet.

638
00:46:52,760 --> 00:47:00,360
All around us are sensors
continuously measuring temperature,
water flow, and ocean currents.

639
00:47:00,360 --> 00:47:06,800
And high in orbit are satellites
busy imaging cloud formations,
forest growth and snow cover.

640
00:47:06,800 --> 00:47:11,360
Scientists speak
of "instrumenting the earth".

641
00:47:13,320 --> 00:47:20,160
And pointing up to the skies
above are powerful new telescopes
mapping the universe.

642
00:47:30,280 --> 00:47:34,760
What's happening in astronomy
is typical of how profoundly

643
00:47:34,760 --> 00:47:39,760
this new torrent of data
is transforming science.

644
00:47:39,760 --> 00:47:45,280
Astronomers are now addressing many
enduring mysteries of the cosmos

645
00:47:45,280 --> 00:47:49,600
by applying statistical methods
to all this new data.

646
00:47:59,800 --> 00:48:03,360
The galaxy is a very big place and
it's got billions of stars in it,

647
00:48:03,360 --> 00:48:09,400
and so to put together a coherent
picture of the whole galaxy requires
having an enormous amount of data.

648
00:48:09,400 --> 00:48:13,720
And before you could do
a large sky survey with
sensitive, digital detectors

649
00:48:13,720 --> 00:48:16,880
that meant that you could map many,
many stars all at once,

650
00:48:16,880 --> 00:48:20,680
it was very difficult to build up
enough data on enough of the galaxy.

651
00:48:24,600 --> 00:48:28,560
In the past, large surveys
of the night sky had to be done

652
00:48:28,560 --> 00:48:32,400
by exposing thousands
of large photographic plates.

653
00:48:32,400 --> 00:48:37,200
But these surveys could take
25 years or more to complete.

654
00:48:39,040 --> 00:48:44,680
Then, in the 1990s, came digital
astronomy and a huge increase

655
00:48:44,680 --> 00:48:49,600
in both the amount
and the accessibility of data.

656
00:48:49,600 --> 00:48:55,960
The Sloan Sky Survey
is the world's biggest yet,
using a massive digital sensor

657
00:48:55,960 --> 00:49:00,840
mounted on the back
of a custom-built telescope
in New Mexico.

658
00:49:00,840 --> 00:49:05,240
It's scanned the sky night
after night for eight years,

659
00:49:05,240 --> 00:49:09,800
building up a composite picture
in unprecedented resolution.

660
00:49:09,800 --> 00:49:14,840
The Sloan is some of the best,
deepest survey data
that we have in astronomy.

661
00:49:14,840 --> 00:49:18,760
Both on our own galaxy and
on galaxies further away from ours.

662
00:49:24,080 --> 00:49:27,320
All the Sloan data
is on the internet,

663
00:49:27,320 --> 00:49:34,120
and with it astronomers
have identified millions of hitherto
unknown stars and galaxies.

664
00:49:34,120 --> 00:49:37,480
They also comb the database
for statistical patterns

665
00:49:37,480 --> 00:49:42,800
which will prove, disprove,
or even suggest new theories.

666
00:49:42,800 --> 00:49:49,160
So we have this idea that galaxies
grow, they become large galaxies like
the one we live in, the milky way,

667
00:49:49,160 --> 00:49:55,880
not all at once, or not smoothly,
but by continuously incorporating,

668
00:49:55,880 --> 00:49:59,160
basically cannibalising,
smaller galaxies.

669
00:49:59,160 --> 00:50:04,000
They dissolve them
and they become part
of the bigger galaxy as it grows.

670
00:50:06,040 --> 00:50:12,520
It's a startling idea,
and, in the Sloan data,
is the evidence to support it.

671
00:50:12,520 --> 00:50:16,280
Groups of stars that came
from cannibalised galaxies

672
00:50:16,280 --> 00:50:21,240
stand out in the Sloan data
as statistically different
from other stars

673
00:50:21,240 --> 00:50:24,280
because they move
at a different velocity.

674
00:50:24,280 --> 00:50:28,680
Each big spike
on one of these distribution graphs

675
00:50:28,680 --> 00:50:35,120
means Professor Rockosi has found
a group of stars all travelling
in a different way to the rest.

676
00:50:35,120 --> 00:50:38,360
They are the telltale
patterns she's looking for.

677
00:50:40,240 --> 00:50:44,960
The evidence is accumulating
that, in fact, this really is
how galaxies grow,

678
00:50:44,960 --> 00:50:47,440
or an important way
in which how galaxies grow.

679
00:50:47,440 --> 00:50:53,000
And so this is an important part
of understanding how galaxies form,
not only ours but every galaxy.

680
00:50:56,360 --> 00:51:00,400
The more data there is,
the more discoveries can be made.

681
00:51:00,400 --> 00:51:03,320
And the technology
is getting better all the time.

682
00:51:03,320 --> 00:51:07,560
The next big survey telescope
starts its work in 2015.

683
00:51:07,560 --> 00:51:10,760
It will leave Sloan in the dust!

684
00:51:10,760 --> 00:51:16,160
Sloan has taken eight years to cover
one quarter of the night sky.

685
00:51:17,680 --> 00:51:25,680
The new telescope will scan
the entire sky, in even greater
resolution, every three days!

686
00:51:34,120 --> 00:51:41,000
The vast amounts of data
we have today allows researchers
in all sorts of fields

687
00:51:41,000 --> 00:51:46,280
to test their theories
on a previously unimaginable scale.

688
00:51:46,280 --> 00:51:53,600
But more than this,
it may even change
the fundamental way science is done.

689
00:51:53,600 --> 00:51:58,560
With the power of today's computers
applied to all this data,

690
00:51:58,560 --> 00:52:03,880
the machines might even be able
to guide the researchers.

691
00:52:14,600 --> 00:52:17,920
We're at a potentially
profoundly important

692
00:52:17,920 --> 00:52:22,560
and potentially one of the most
significant points in science,

693
00:52:22,560 --> 00:52:24,680
and certainly one of
the most exciting,

694
00:52:24,680 --> 00:52:32,080
where the potential to transform
not just how scientists do science
but even what science is possible.

695
00:52:32,080 --> 00:52:34,680
And what will power
that transformation

696
00:52:34,680 --> 00:52:38,400
of both how science is done
and even what science is possible

697
00:52:38,400 --> 00:52:40,120
is going to be computation.

698
00:52:41,800 --> 00:52:49,440
Many of the dynamics of the natural
world, like the interplay between
the rainforests and the atmosphere,

699
00:52:49,440 --> 00:52:53,560
are so complex that we don't
as yet really understand them.

700
00:52:53,560 --> 00:52:59,280
But now computers are generating
literally tens of thousands
of different simulations

701
00:52:59,280 --> 00:53:03,480
of how these
biological systems might work.

702
00:53:03,480 --> 00:53:07,840
It's like creating thousands
of hypothetical parallel worlds.

703
00:53:07,840 --> 00:53:10,640
Each and every one
of these simulations

704
00:53:10,640 --> 00:53:18,360
is analysed with statistics
to see if any are a good match
for what is observed in nature.

705
00:53:18,360 --> 00:53:21,840
The computers can now
automatically generate,

706
00:53:21,840 --> 00:53:26,240
test and discard hypotheses
with scarcely a human in sight.

707
00:53:28,240 --> 00:53:35,120
This new application of statistics
will become absolutely vital
for the future of science.

708
00:53:35,120 --> 00:53:39,400
It's creating a new paradigm,
if you like,

709
00:53:39,400 --> 00:53:42,640
in science, in the way
in which we can do science,

710
00:53:42,640 --> 00:53:45,280
which is increasingly...

711
00:53:45,280 --> 00:53:51,160
Which one might characterise as...
data-centric or data driven

712
00:53:51,160 --> 00:53:55,000
rather than being hypothesis-driven
or experimentally-driven.

713
00:53:55,000 --> 00:53:58,240
So, it's exciting times
in terms of the science,

714
00:53:58,240 --> 00:54:02,200
in terms of the computation
and in terms of the statistics.

715
00:54:08,800 --> 00:54:15,480
Now, if all that sounds a bit
abstract and theoretical to you,
how about one final frontier?

716
00:54:15,480 --> 00:54:19,040
Could statistics even make
sense of your feelings?

717
00:54:21,200 --> 00:54:25,800
In California - where else? -
one computer scientist

718
00:54:25,800 --> 00:54:32,680
is harvesting the internet to try
to divine the patterns of our
innermost thoughts and emotions.

719
00:54:44,800 --> 00:54:46,360
This is the madness movement.

720
00:54:46,360 --> 00:54:50,960
The madness movement represents
a skyscraper view of the world.

721
00:54:50,960 --> 00:54:54,880
Each of these brightly coloured dots
is an individual feeling

722
00:54:54,880 --> 00:54:58,720
expressed by someone out there
in a blog or a tweet.

723
00:54:58,720 --> 00:55:04,480
And when you click on the dot
it explodes to reveal the
underlying feeling of that person.

724
00:55:04,480 --> 00:55:07,080
This is what people say
they're feeling today.

725
00:55:07,720 --> 00:55:10,160
Better...safe...

726
00:55:10,160 --> 00:55:12,040
crappy...

727
00:55:12,040 --> 00:55:14,560
well...

728
00:55:14,560 --> 00:55:18,440
pretty...special...

729
00:55:18,440 --> 00:55:20,800
sorry...alone...

730
00:55:25,560 --> 00:55:29,040
So, every minute, We Feel Fine
crawls the world's blogs,

731
00:55:29,040 --> 00:55:34,120
takes all the sentences
that start with the words
"I feel" or "I am feeling",

732
00:55:34,120 --> 00:55:35,920
and puts them in a database.

733
00:55:35,920 --> 00:55:40,080
We collect all the feelings
and we count the most common.

734
00:55:40,080 --> 00:55:43,320
They are better...bad...

735
00:55:43,320 --> 00:55:45,640
good...right...

736
00:55:45,640 --> 00:55:48,520
guilty...sick...

737
00:55:48,520 --> 00:55:51,680
the same...like shit...

738
00:55:51,680 --> 00:55:54,720
sorry...well...

739
00:55:54,720 --> 00:55:56,240
and so on.

740
00:55:58,320 --> 00:56:01,760
And we can take a look at any
one feeling and analyse it.

741
00:56:01,760 --> 00:56:04,800
Right now a lot of people
are feeling happy.

742
00:56:04,800 --> 00:56:11,320
We can take a look at all the
people who are happy and break it
down by age, gender or location.

743
00:56:11,320 --> 00:56:16,840
Since bloggers have public profiles
we have that information and
so we can ask questions like,

744
00:56:16,840 --> 00:56:21,400
"Are women happier than men?"
or, "Is England happier
than the United States?"

745
00:56:30,240 --> 00:56:33,120
We find that, as people get older,
they get happier.

746
00:56:33,120 --> 00:56:40,560
And, moreover, we find that
for younger people they associate
happiness more with excitement,

747
00:56:40,560 --> 00:56:47,000
and, as people get older,
they associate happiness
more with peacefulness.

748
00:56:51,240 --> 00:56:57,760
And we also find that women feel
loved more often than men,
but also more guilty.

749
00:56:57,760 --> 00:57:02,480
While men feel good more often
than women, but also more alone.

750
00:57:06,640 --> 00:57:12,480
As people lead more and
more of their lives online,
they leave behind digital traces,

751
00:57:12,480 --> 00:57:19,840
and with these digital traces
we can begin to statistically analyse
what it means to be human.

752
00:57:51,280 --> 00:57:54,480
So where does all of this leave us?

753
00:57:54,480 --> 00:58:00,160
We generate unimaginable
quantities of data
about everything you can think of.

754
00:58:00,160 --> 00:58:02,800
We analyse it to reveal
the patterns.

755
00:58:02,800 --> 00:58:10,480
And now not only experts
but all of us can understand
the stories in the numbers.

756
00:58:18,160 --> 00:58:21,080
Instead of being
led astray by prejudice,

757
00:58:21,080 --> 00:58:28,160
with statistics at our fingertips,
our eyes can be open
for a fact-based view of the world.

758
00:58:28,160 --> 00:58:33,760
So, more than ever before, we can
become authors of our own destiny.

759
00:58:33,760 --> 00:58:36,800
And that's pretty
exciting isn't it?!

760
00:58:37,680 --> 00:58:44,200
# 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11,
12, 13, 14, 15, 16, 17, 18, 19, 20

761
00:58:44,200 --> 00:58:50,800
# 1, 22, 3, 24, 25, 26, 27, 28, 9,
30, 31, 32, 3, 34, 35, 36, 7

762
00:58:50,800 --> 00:58:54,440
# 38, 39, 40, 41, 42, 3,
44, 45, 46, 47

763
00:58:54,440 --> 00:58:58,680
LYRICS DEGENERATE INTO GIBBERISH

764
00:59:08,680 --> 00:59:13,400
GIBBERISH DEGENERATES INTO NOISE

765
00:59:13,400 --> 00:59:14,440
# 100. #